A memory device typically comprises a semiconductor die including a plurality of memory cells arrayed in rows and columns. The die also includes input/output terminals configured to transmit information relating to a cell's address, the row and column that may be used to access that cell. The die further includes at least one input/output terminal configured to transmit data to or from the cell at a certain address. At least one additional input configures the die to receive external data at the data terminal and store that data in a cell designated by the address signal. In such a configuration, the memory is in a “write” mode. Changing the state of that input (or providing another input) configures the die to transmit to the data terminal data stored in a cell designated by the address signal. In such a configuration, the memory is in a “read” mode.
It may be desirable to test the memory cells at some point during the construction of a memory device. One conventional method of testing the memory cells is to write data to memory cells of the die, then read data from the memory cells, and compare the input with the output. Regardless of the exact test strategy, comparing the written data to the read data may result in instances where such data does not match for particular cells. Such a mismatch suggests that those cells may contain defects that may prevent proper operation of the memory.
In order to restore the functionality of a die having at least one failed memory cell, redundant cells provided on the die may be used to replace the failed cells. For example, given a column of main memory cells, including a failed cell, it is known to replace the entire column with a redundant column of cells. It may be appreciated that such a repair strategy devotes redundant cells to replacing main cells that may have not failed a test. Hence, alternative repair strategies involve using only a segment of a redundant column to replace the corresponding number of cells in a main memory column. Because a whole redundant column has not been committed to repairing what may be only a single defective cell, the redundant segments saved from use in a particular repair may be available to replace other similarly sized portions of main memory that contain failed cells.
Testing, in turn, may be enacted in a manner that accommodates the segmentation of redundancy. Specifically, only rows served by a particular redundant segment may be activated at one time during testing. Once testing and repair of those rows is completed, they are deactivated, and another section of rows, served by another redundant segment, is activated. The activation and deactivation of rows by sections may add to test time despite the desire in the art to shorten test times.
Returning to the topic of repair, circuitry may be provided on the die that allows a redundant element, be it a redundant cell, a segment of a redundant column or row, a full redundant column or row, etc., to replace a corresponding number of main memory cells that include the failed memory cell. Such circuitry may include programmable devices, wherein programming those devices indicates not only that a redundant element is being used but also which address or addresses are now to be associated with the redundant element. The programmable devices may be anti-fuses, i.e., capacitive devices that may be closed or blown by breaking down a dielectric layer therein with a relatively high voltage. In such a case, a specific combination of anti-fuses may be programmed to correspond to an address of a column having at least one failed memory cell. Such anti-fuses may be arranged in a bank with the number of anti-fuses corresponding to the number of address bits for a column address. Thus, if the failed column has a 2-bit binary address, then two anti-fuses in a bank may be programmed to store this address. A die may contain several anti-fuse banks for a respective number of redundant columns. In addition, enable circuitry including an anti-fuse may be included in each bank, wherein programming the enable circuitry's anti-fuse indicates or allows the use of a redundant element.
After repair, the incoming addresses are compared to the addresses programmed into the anti-fuse banks. If a match is detected, then a match signal is transmitted to control circuitry, such as a column decoder. The match signal indicates that a redundant column should be accessed and the failed column in the main memory array should be ignored. The control circuitry directs data accordingly.
It follows that it is desired to refrain from using defective redundant cells to repair failed main memory cells. Accordingly, it is known in the art to test the redundant cells in a manner that may be similar to testing the main memory cells. During one particular stage of testing, e.g., before the die is completely packaged, such a test may be conducted, defective redundant cells may be identified, the location of such cells may be stored by the test device, and the test device may refrain from using those redundant cells in repairing the die. However, prior art techniques tend to not retain the association between a particular die and the test device's information concerning that die's failed redundant cells. As a result, later testing, such as testing after packaging in a different test device, will not benefit from that data, and repairs resulting from that later testing may in fact undesirably use failed redundant elements.